THE QUESTION WE STARTED WITH
What must humans learn when answers become abundant?
Personalised support should strengthen independent thinking. A polished answer with a tutor does not show what a pupil can explain alone.
THE NEXT QUESTION
What can students explain and solve independently a month after using an AI tutor, and which children benefit least?
Greece’s education ministry describes a pilot in 20 lyceums. The evidence we need next concerns independent learning, support for pupils with different needs and the teacher’s actual workload.
Organizing a discussion on this question?
Use these findings and sources as the starting point for a panel, interview, email exchange or small roundtable.
- 01The starting point
Personalised support should strengthen independent thinking. A polished answer with a tutor does not show what a pupil can explain alone.
- 02Further evidence
Bastani’s school experiment separates better assisted practice from weaker subsequent unaided performance; a tutor shaped by teachers largely mitigated that harm. The physics tutor designed by Kestin’s team improved immediate learning. Tutor design and independent reasoning, rather than polished homework, become the test.
Still open These short studies involve different ages, subjects and settings. They do not establish lasting gains, reliable Greek explanations or the full workload in Greek schools.
- 03The decision to test
For one lesson, agree what help pupils may use, how they will explain the answer without it and how learning will be checked later. Include an accessible alternative and all the time the teacher needs. The test is whether the teaching approach helps pupils learn, rather than how polished the assisted work looks.
Still open Delayed transfer, effects for different pupil needs and sustainable staffing remain open. No new Greek pilot result or national success threshold is available here.
WHO CAN HELP ANSWER THIS?
The people closest to the problem.
Teacher in a Greek pilot school
What must a student demonstrate unaided before you conclude that learning improved?
An initial task, an unfamiliar independent task one month later and the marking rubric.Student and parent
When did the tutor help you understand, and when did you accept an answer you could not explain?
One worked example, a detected mistake and an explanation in the student’s own words.School leader or pilot evaluator
Who gets access and support, and what finding would change the programme?
Participation by need, teacher preparation and correction time, language checks and publication date for results.Each perspective can fill a different gap in what we know.
Follow the question across the network.
ThoughtLeaders.gr The attributed positions ↗AIStrategy.gr The question in this sector ↗EventChanger The conversation ↗Sources and what remains uncertain
Read the original work alongside the analysis. For the local case, we still need the evidence requested above.
- Sal KhanHow AI could save (not destroy) education ↗April 2023 · khan-ted-202304
- Michalis BletsasInterview on AI and learning on ERTnews Syndeseis ↗26 August 2026 · bletsas-learning-2026
- Hamsa Bastani, Osbert Bastani, Alp Sungu, Haosen Ge, Özge Kabakcı, Rei MarimanGenerative AI without guardrails can harm learning: Evidence from high school mathematics ↗25 June 2025 · bastani-learning-20250625
- Greg Kestin, Kelly Miller, Anna Klales, Timothy Milbourne, Gregorio PontiAI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting ↗3 June 2025 · kestin-tutoring-20250603